Mild cognitive impairment screening method, computer program accompanying the method, and screening device

Abstract

The present application provides a mild cognitive impairment screening method, a computer program accompanying the method, and a screening device, which can easily and quickly perform high-precision MCI screening without using a complex, large, and expensive device and without requiring professional knowledge. The mild cognitive impairment screening method of the present application includes: a measurement step (S1) of measuring opening and closing movement of two fingers, i.e., finger tapping movement; a calculation step (S3) of calculating a predetermined characteristic quantity based on measurement data obtained by the measurement step in a predetermined time during right-hand finger tapping only and / or left-hand finger tapping only and / or simultaneous two-hand finger tapping and / or alternating two-hand finger tapping; and a determination step (S5) of performing determination related to mild cognitive impairment based on the characteristic quantity calculated in the calculation step (S3). In the determination step (S5), a discrimination model is generated by machine learning including artificial intelligence (step S4), and determination related to mild cognitive impairment is performed using the generated discrimination model.

Description

Technical Field

[0001] This invention relates to a method for screening mild cognitive impairment, and more particularly to a method for measuring finger tapping movements and using the measurement results to screen for mild cognitive impairment, a computer program accompanying the method, and a screening device. Background Technology

[0002] With the aging of society, the number of patients with Alzheimer's dementia is increasing year by year. If it can be detected early, the progression of the disease can be slowed down with medication. Sometimes it is difficult to distinguish between age-related symptoms such as forgetting things and the disease itself, and many cases only seek medical attention after the condition has become severe.

[0003] In this context, screening tests for the early detection of Alzheimer's disease have traditionally involved blood tests, olfactory tests, and tests simulating doctor's consultations on tablets. However, these methods are burdensome for test takers due to pain during blood collection and lengthy testing times. On the other hand, tests that are less burdensome for test takers, such as single-handed finger movement measurements using tablets based on button presses, assess cognitive function. However, these methods suffer from limitations in achieving sufficient accuracy. If a high-precision, low-burden, and simple screening test could be developed, it would facilitate the early detection of Alzheimer's disease, improve patients' quality of life, and reduce medical and nursing care costs.

[0004] In contrast, recent years have clarified that specific motor patterns in Alzheimer's dementia can be extracted from the opening and closing movements of the thumbs and index fingers (finger tapping movements), and a high correlation has been confirmed with finger motor measurements and dementia examinations based on general medical history. These findings can be seen as capturing the decline in rhythmic motor function of the fingers caused by brain atrophy in Alzheimer's dementia through finger tapping movement measurements. Furthermore, the fingers can be considered a second brain, as many areas of the brain are related to finger movements. These movements are not limited to Alzheimer's dementia but are also associated with cerebrovascular dementia, Lewy micropenis, Parkinson's disease, and developmental coordination disorders (such as inability to hop on one leg or jump rope). In other words, the state of the brain can be inferred from finger tapping movements. Moreover, by using finger tapping movements as a "measuring stick" for brain health, fine motor function of the fingers can be quantified, thus enabling applications in various fields such as healthcare, rehabilitation, and daily living assistance.

[0005] Furthermore, as a method for high-precision measurement and evaluation of finger tapping movements, for example, Patent Document 1 discloses a motor function evaluation system, which includes a motor function measuring device and an evaluation device. The motor function measuring device calculates motion data based on the relative distance between paired transmitting and receiving coils installed on the movable part of the organism, and the evaluation device evaluates the organism's motor function based on the motion data received from the motor function measuring device. Specifically, Patent Document 1 shows the following: by using a magnetic sensor worn on the fingertip, the change in magnetic force due to the tapping movement of two fingers is converted into an electrical signal, and the movement is measured and quantified to capture characteristic quantities representing the characteristics of finger movements, thereby obtaining the state of brain function.

[0006] Existing technical documents

[0007] Patent documents

[0008] Patent Document 1: Japanese Patent Application Publication No. 2016-49123 Summary of the Invention

[0009] The problem that the invention aims to solve

[0010] As mentioned above, early detection of dementia is extremely important. To this end, it is also necessary to identify the early stage of dementia, namely mild cognitive impairment (MCI), which signifies a decline in subjective cognitive function.

[0011] Previously, screening for MCI involved diagnosis by a dementia specialist, comprehensive geriatric evaluation (CGA) by a clinical psychologist, psychological tests (FAB, RCPM, ADAS, etc.), magnetic resonance imaging (MRI), single-photon computed tomography or electrocardiogram, and blood tests. Based on these results, the final decision was made by a dementia specialist.

[0012] However, MCI screening based on such methods comes with a variety of problems: the examination is invasive, the examination time is long, expertise is required, and complex, large and expensive equipment must be used.

[0013] The present invention was made in view of the above circumstances, and its purpose is to provide a method for screening mild cognitive impairment, a computer program accompanying the method, and a screening device that can easily and quickly perform high-precision MCI screening without using complex, large and expensive devices or requiring professional knowledge.

[0014] Methods for solving problems

[0015] To address the aforementioned issues, this invention provides a method for screening mild cognitive impairment, characterized by comprising a measurement step, a calculation step, and a judgment step. In the measurement step, the opening and closing movements of two fingers, i.e., finger tapping movements, are measured. In the calculation step, a predetermined feature quantity is calculated based on the measurement data obtained within a predetermined time period during the measurement steps when the right hand's fingers tap only, / or the left hand's fingers tap only, / or both hands tap simultaneously, and / or both hands tap alternately. In the judgment step, a judgment related to mild cognitive impairment is made based on the feature quantity calculated in the calculation step. In this judgment step, a discriminant model is generated using machine learning incorporating artificial intelligence, and this generated discriminant model is used to make the judgment related to mild cognitive impairment.

[0016] The inventors conducted clinical research focusing on finger tapping movements and successfully extracted movement patterns unique to individuals with mild cognitive impairment (MCI) from the tapping movements of the thumbs and index fingers. Furthermore, the inventors gained the insight that by using a technique that quantifies fine finger movements with magnetic sensors and processing various feature quantities of the repetitive opening and closing of the two fingers using machine learning incorporating artificial intelligence, high-precision screening for mild cognitive impairment can be achieved. Moreover, based on this achievement, the present invention provides a screening method to support the early detection of MCI, namely, a mild cognitive impairment screening method.

[0017] Specifically, in machine learning, data from healthy individuals and individuals with MCI are provided as information regarding right-handed finger-tapping movements and / or left-handed finger-tapping movements and / or simultaneous finger-tapping movements of both hands and / or alternating finger-tapping movements of both hands. Based on this information, a discriminant model is generated using a predetermined algorithm incorporating artificial intelligence to distinguish between healthy individuals and those with MCI (classifying them as MCI individuals). Then, this generated discriminant model is used to make a determination related to MCI. More specifically, in this machine learning, a pre-prepared healthy group and an MCI group are input into a machine learning algorithm capable of discriminating between the two groups (described later) to generate a discriminant model. Then, when new subject data is input, the determination result is output.

[0018] According to the above-described structure of the present invention, since the massive amount of data, which consists of a wide variety of features, is processed through machine learning incorporating artificial intelligence, no specialized knowledge is required, and MCI screening can be performed easily and with high accuracy in a short time. Furthermore, the finger tapping measurement itself can be easily performed non-invasively using a simple and compact device, thus enabling measurement and evaluation with less burden on the subject in a shorter measurement time.

[0019] It should be noted that in the above configuration, the measurement step of measuring the opening and closing movements of two fingers, i.e., finger tapping movements, can use any measurement method as long as it can measure finger tapping movements. For example, it could be a measurement method using magnetic sensors worn on the two fingers, or a measurement method that processes image data obtained by capturing finger movements, or a measurement method based on detecting finger movements by tapping a touch panel. Furthermore, in the above configuration, the "determination related to mild cognitive impairment" includes determining whether it is MCI, determining whether the probability of MCI is high or low, the probability of having MCI, and indicating the degree of MCI (determination), etc.

[0020] Furthermore, in the above structure, it is preferable to make a determination related to MCI based on a number of feature quantities obtained through various finger tapping movements, especially the feature quantities in which significant differences are observed between healthy individuals and MCI individuals.

[0021] In addition to the above-described method for screening mild cognitive impairment (e.g., by processing by a computer according to a predetermined algorithm), the present invention also provides a computer program for executing the method and a device for screening mild cognitive impairment capable of executing the above-described method.

[0022] Invention Effects

[0023] The mild cognitive impairment screening method, computer program, and apparatus of the present invention do not use complex, large, and expensive devices, do not require professional knowledge, and can easily and quickly perform high-precision MCI screening. Attached Figure Description

[0024] Figure 1 This is a block diagram illustrating the schematic configuration of a mild cognitive impairment screening device according to one embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the hands of a subject who wore tapping sensors on their thumb and index finger.

[0026] Figure 3 It means Figure 1 A flowchart illustrating an example of a mild cognitive impairment screening method according to one embodiment of the present invention, which can be performed by a mild cognitive impairment screening device.

[0027] Figure 4 This is a table representing one example of the number of features required to determine mild cognitive impairment screening.

[0028] Figure 5 This is a table showing examples of machine learning algorithms (machine learning models) that distinguish between the healthy group and the MCI group.

[0029] Figure 6This is a table representing the discrimination accuracy (F1 score of the top 5%) between the MCI group and the healthy group for individual feature quantities.

[0030] Figure 7 This is a table showing the discrimination results of the healthy group and the MCI group based on various machine learning models; it is a rough diagram. Detailed Implementation

[0031] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In this embodiment, by providing the technology shown below, a highly advanced technology is used to contribute to the development of medicine and the realization of a healthy society. The implementation of this mild cognitive impairment screening method (device and computer program) contributes to the United Nations' Sustainable Development Goals (SDGs) "9. Building the foundation for industrial and technological innovation".

[0032] Furthermore, while the following embodiments describe a method (device) for screening mild cognitive impairment, the present invention can also be configured as a computer program capable of performing the processes executed by the method (device) for screening mild cognitive impairment via a computer.

[0033] Figure 1 The figure shows a schematic configuration of a mild cognitive impairment screening device 1 according to one embodiment of the present invention. As shown, the mild cognitive impairment screening device 1 includes a measurement unit 10 and a processor 30. The measurement unit 10 has a tapping sensor 2 that performs magnetic detection on the opening and closing movements of two fingers, i.e., finger tapping movements. The processor 30 processes the measurement data measured by the measurement unit 10.

[0034] The measurement unit 10 measures finger motion data based on the relative distance between the paired transmitting and receiving coils installed on the finger (or other movable part) of the organism. For example, it can detect information on the finger motion of the subject in a time sequence and obtain the subject's motion information as time sequence data (waveform data) related to at least one of distance, velocity, acceleration, and jerk (the value obtained by differentiating acceleration over time).

[0035] The measuring unit 10 includes a tap sensor 2, first and second switching circuits 4 and 5, an AC generator 6 for generating AC, an amplification / filtering circuit 7, an A / D converter 8, a detector 9, a downsampler 10 for downsampling, and a controller 11 for controlling their operation.

[0036] The tap sensor 2 consists of paired transmitting coils 2A (2A') and receiving coils 2B (2B') (or multiple columns of paired coils), for example... Figure 2As shown, the fingers (e.g., fingertips) of the subject's hand 100 (100A, 100A') are fitted with something such as double-sided tape or a fastening strap. Specifically, in Figure 2 In this experiment, a paired transmitting coil 2A and a receiving coil 2B are worn on the thumb 100a and index finger 100b of the subject's right hand 100A, respectively, and a paired transmitting coil 2A' and a receiving coil 2B' are worn on the thumb 100a and index finger 100b of the subject's left hand 100A', respectively (the fingers can be reversed or other fingers can be used). In this case, the transmitting coil 2A (2A') transmits a magnetic field, and the receiving coil 2B (2B') receives (detects) the magnetic field transmitted by the transmitting coil 2A (2A').

[0037] The transmitting coil 2A (2A') is connected to an alternator 6 via a first switching circuit 4. Through the switching action of the first switching circuit 4, an alternating current (e.g., a 20kHz current) from the alternator 6 flows sequentially through the transmitting coil 2A (2A'), generating an alternating magnetic field in the transmitting coil 2A (2A') through which the alternating current flows. The alternator 6 generates an alternating current of a predetermined frequency, and the timing of the current flow is controlled by the controller 11. It should be noted that the signal generated by the alternator 6 is used as the reference signal for the detection operation of the detector 9.

[0038] The controller 11 generates a synchronization signal for controlling the first switching circuit 4 and the second switching circuit 5. With this synchronization signal, the first switching circuit 4 and the second switching circuit 5 can switch simultaneously, operating sequentially for each paired transmitting coil 2A (2A') and receiving coil 2B (2B').

[0039] Furthermore, the receiving coil 2B (2B') is connected to the amplification / filtering circuit 7 via the second switching circuit 5. The output signal from the amplification / filtering circuit 7 is converted into a digital signal by the A / D converter 8, and this digital signal is transmitted to the detector 9. It should be noted that by digitizing the analog data by the A / D converter 8, subsequent processing (downsampling, etc.) becomes easier. In addition, the detector 9 also performs processing to remove the AC magnetic field waveform (noise portion) of a predetermined period after switching by the second switching circuit 5 from the AC magnetic field waveform detected by the receiving coil 2B (2B').

[0040] Furthermore, the timing of the deletion process in the AC magnetic field waveform of each receiving coil 2B (2B') is precisely controlled by the controller 11. After this deletion process, the detector 9 performs full-wave rectification and filtering (mainly based on a low-pass filter (LPF)) using the aforementioned reference signal. Finally, the digital signal processed by the detector 9 is converted (downsampled) by the downsampler 10 into coarse data at a sampling frequency (e.g., 200Hz) that is approximately 1 / 1000th (a predetermined ratio) of the sampling frequency (e.g., 200kHz) in the A / D converter 8. This reduces the overall data volume. Therefore, the output signal can be transmitted at high speed as data from multiple receiving coils even when communication capacity is limited. That is, the amount of data received by the communication interface 12 of the measurement unit 10 from the downsampler 10 is small, so the finger movement data associated with multiple receiving coils can be transferred to the processor 30 at once via wireless or wired (through the communication interface 31 of the processor 30).

[0041] The processor 30, which processes the measurement data measured by the measurement unit 10, includes an arithmetic circuit 33 and a determination circuit 34. The arithmetic circuit 33 calculates a predetermined number of feature quantities within a predetermined time period based on the detection information detected by the tapping sensor 2. Specifically, it calculates the feature quantities or their quantity based on the measurement data output from the measurement unit 10 (measured by the tapping sensor 2 for each group of healthy individuals and MCI patients) showing finger tapping movements of only the right hand and / or only the left hand and / or simultaneous finger tapping movements of both hands and / or alternating finger tapping movements of both hands. (The selection of these feature quantities or their quantity can be performed automatically based on a predetermined algorithm, or it can be based on instructions input through the operation input interface 38 described later.) The determination circuit 34 makes a determination related to mild cognitive impairment based on the feature quantities calculated by the arithmetic circuit 33. Here, in this embodiment, the determination circuit 34 generates a discriminant model using machine learning incorporating artificial intelligence, and uses this generated discriminant model to make a determination related to mild cognitive impairment. In addition, the processor 30 also includes an adjustment circuit 40, which adjusts the optimal values ​​of the hyperparameters in the machine learning of the decision circuit 34 so that the F1 score (F1 value) used as an evaluation metric in the machine learning model is 0.7 or higher.

[0042] One example of the feature quantities used in the determination related to MCI. Figure 4 As shown and listed below.

[0043] 1. The maximum amplitude of the distance between two fingers.

[0044] 2. Total traveling distance of the two fingers.

[0045] 3. Average of the local maximum distances between two fingers.

[0046] 4. Standard deviation of the local maximum distance (SD of the distance between two fingers)

[0047] 5. The slope of the approximate line of the local maximum point (the approximate curve is the regression line when time is plotted on the horizontal axis and the maximum value on the vertical axis).

[0048] 6. The coefficient of variation of the local maximum distance (CV).

[0049] (The coefficient of variation is the value obtained by dividing the standard deviation by the mean.)

[0050] 7. Standard deviation of local amplitude (SD of local maximum distance in three adjacent taps), which is the standard deviation of the maximum distance between two fingers in three adjacent taps.

[0051] 8. Maximum velocity amplitude, which is the maximum velocity of the fingers during the opening and closing motion of two fingers.

[0052] 9. The average of the local maximum velocity points when the two fingers are released.

[0053] 10. Average of local maximum velocity points during the closing motion of two fingers.

[0054] 11. Standard deviation of the local maximum velocity (SD of the two fingers when releasing the finger)

[0055] 12. Standard deviation of the local minimum velocity (SD) of the finger velocity during the closing motion of two fingers.

[0056] 13. Energy balance is the ratio of the sum of the squares of the velocities of the two fingers when they are released to the sum of the squares of the velocities of the two fingers when they are closed.

[0057] 14. Total energy, which is the sum of the squares of the velocities over the entire measurement period.

[0058] 15. The coefficient of variation of the local maximum velocity (CV) of the finger's velocity at the point of maximum velocity when the two fingers are released.

[0059] 16. The coefficient of variation of the local minimum velocity (CV) of the finger during the closing motion of two fingers.

[0060] 17. Number of freezing points calculated from velocity: This is the number obtained by subtracting the number of large-scale finger taps from the number of reciprocating strokes when the velocity waveform changes from positive to negative.

[0061] 18. The average of the distance ratio of the velocity peak in extending movement, which is the average of the ratios related to the distance at the maximum velocity during the two-finger release movement, and which is related to the ratio assuming the amplitude of the finger tapping is set to 1.0.

[0062] 19. The average of the distance rate of velocity peak in flexing movement, which is the average of the distance ratio related to the distance at the minimum velocity in the two-finger closing movement, and related to the ratio with the amplitude of the finger tapping set to 1.0.

[0063] 20. The ratio of distance rates of velocity peak in extending and flexing movements to the distance ratio of the velocity peak in extending and flexing movements.

[0064] 21. The standard deviation of the distance rate of velocity peak in extending movement (SD of the distance rate of velocity peak in extending movement).

[0065] 22. Standard deviation of the distance rate of velocity peak in flexing movement (SD of the distance rate of velocity peak in flexing movement)

[0066] 23. The maximum amplitude of acceleration of the fingers during the closing motion of two fingers.

[0067] 24. The average of the local maximum acceleration in extending movement.

[0068] 25. The average of the local minimum acceleration in extending movement.

[0069] 26. The average of the local maximum acceleration in flexing movement (the average of the local maximum acceleration points in flexing movement).

[0070] 27. Average of local minimum acceleration in flexing movement (mean of the local minimum acceleration point during the two-finger closing motion).

[0071] 28. Average of contact duration between the two fingers during tapping.

[0072] 29. Standard deviation of contact duration (SD of contact time) between the two fingers during tapping.

[0073] 30. The coefficient of variation of contact duration between the two fingers during tapping (CV of contact duration).

[0074] 31. The number of zero crossover points of acceleration during tapping, i.e., the average number of positive and negative changes in acceleration during one cycle of finger tapping.

[0075] 32. Number of freezing points calculated from acceleration: This is the value obtained by subtracting the number of large-scale opening and closing finger taps from the number of reciprocating strokes during one cycle of finger tapping when the sign of acceleration changes.

[0076] 33. Number of taps

[0077] 34. The time interval between each strike is the average of the strike intervals.

[0078] 35. Frequency of taps (average tap interval)

[0079] 36. Standard deviation of inter-tapping interval (SD of inter-tapping interval)

[0080] 37. Coefficient of variation of inter-tapping interval (CV of inter-tapping interval)

[0081] 38. Inter-tapping interval variability, which is the cumulative value of the frequency range of 0.2~2.0Hz when performing spectral analysis on the tapping interval.

[0082] 39. Skewness of inter-tapping interval distribution, which is the degree to which the frequency distribution of the tapping interval is skewed compared to the normal distribution.

[0083] 40. Standard deviation of inter-tapping interval in three adjacent taps (SD of inter-tapping interval in three adjacent taps).

[0084] 41. The average of the phase difference between the left and right tapping (the phase difference between the tapping of the fingers of the right hand and the tapping of the fingers of the left hand).

[0085] 42. The standard deviation of the phase difference between the left and right tapping (SD of the phase difference between the tapping of the right and left hands).

[0086] 43. Similarity of hands (multiple hands similarity), which represents the correlation value when the time deviation is 0, given the cross-correlation function applied to the waveforms of the left and right hands.

[0087] 44. The time lag of similarity of hands (the time lag of similarity among multiple hands) represents the value of the correlation calculated in the previous term that has the largest time lag.

[0088] Of the 44 features listed above, features 1-40 are obtained through tapping motions using only the right or left hand, while features 41-44 are obtained through simultaneous or alternating tapping motions using both hands. When using only one hand, 40 features for each of items 1-40 are obtained, resulting in a total of 40 × 2 = 80 features for both hands. Furthermore, when using both hands simultaneously and alternately, 40 × 2 = 80 features are obtained for each of items 1-40 on both hands, and 4 features are obtained for items 41-44 on both hands, totaling 84 features. Therefore, a total of 84 × 2 = 168 features are obtained for both simultaneous and alternating tapping motions. The result is that for each subject, a total of 80 + 168 = 248 features are obtained.

[0089] The determination of association with mild cognitive impairment can be made based on all of these features, but in particular in this embodiment, the determination of association with mild cognitive impairment is made based on the following features among these 248 features that show significant differences between healthy individuals and those with MCI (high correlation with healthy individuals and those with MCI).

[0090] That is, in this embodiment, the following feature quantities are used in the determination of mild cognitive impairment: for finger tapping movements of only the left hand, feature quantities of items 4, 12, 15, 16, 21, 25, 27, 28, 29, 32, 34, 36, 38, and 40 are used; for finger tapping movements of only the right hand, feature quantities of items 16, 25, 27, 28, 29, 34, and 38 are used; for finger tapping movements of both hands simultaneously, feature quantities of items 12, 15, 18, 25, 27, and 28 are used for the left hand. The characteristic quantities of items 8, 29, 34, 38 and right-hand items 18, 20, 25, 28, 29, 33, 34, 35, 38, regarding the alternating finger tapping motion of both hands, using left-hand items 17, 18, 20, 21, 22, 25, 28, 29, 30, 31, 32, 34, 36, 37, 38, 40 and right-hand items 3, 20, 25, 27, 28, 29, 31, 32, 34, 36, 37, 38, 40 (hereinafter referred to as selected characteristic quantities).

[0091] In addition, the mild cognitive impairment screening device 1 also includes a display 37 that displays various data (such as calculation results calculated by the arithmetic circuit 33) including the judgment results of the judgment circuit 34 of the processor 30, a memory 36 that stores the above-mentioned various data, and an operation input interface 38 that can input the required data and commands to the processor 30 through operation.

[0092] In the above configuration, the processor 30 is composed of a CPU, etc., and performs the operation control processing of the various circuits 33, 34, 40 by executing the operating system (OS) and various motion control applications stored in the memory 36, and controls the startup of various applications.

[0093] The memory 36 is composed of flash memory and other components, and stores programs such as operating systems, image, sound, document, display, and measurement applications for motion control. In addition, the memory 36 stores basic data required for the basic operations of the operating system, file data used in various applications, and other information data.

[0094] It should be noted that the processing in processor 30 can also be stored as an application, and the measurement and processing of finger movements, as well as the calculation and analysis of various feature quantities, can be performed by launching the application. Alternatively, the measurement results can be received from the information processing terminal and the feature quantities can be calculated and analyzed in an external server device or similar device with high computing performance and large capacity.

[0095] In addition, the operation input interface 38 generally uses input units based on keyboards, buttons, touch keys, etc., but gesture operation and voice input can also be used, for example, to set the information that the input subject should input.

[0096] Furthermore, the communication interface 31 not only receives measurement results from the measurement unit 10, but can also communicate wirelessly with server devices located in other locations via short-range wireless communication, wireless LAN, or base station communication. In this case, measurement data and analyzed and calculated characteristic quantities can also be transmitted and received with server devices via the transceiver antenna 39 during wireless communication. It should be noted that short-range wireless communication can be achieved using electronic tags, but is not limited to this. As long as wireless communication is possible in the vicinity of other information terminals, wireless LANs such as Bluetooth, IrDA (Infrared Data Association), Zigbee, HomeRF (Home Radio Frequency), or Wi-Fi can also be used. Additionally, for base station communication, long-range wireless communication such as W-CDMA (Wideband Code Division Multiple Access) and GSM (Global System for Mobile communications) can be used. It should also be noted that Ultra Wide Band (UWB) wireless systems can be used to detect the positional relationship and orientation between terminals. Although not shown, the communication interface 31 can also use other methods such as optical communication based on acoustic waves as a form of wireless communication. In this case, instead of the transceiver antenna 39, an optical emission / receiving unit and an acoustic wave output / input interface are used respectively.

[0097] It should be noted that in this embodiment, the measuring unit 10 and the processor 30 each have the above-mentioned constituent elements, but they may also have a functional unit that combines at least some or all of these constituent elements. In short, as long as the function of these constituent elements is ensured, it can be any configuration.

[0098] The mild cognitive impairment screening device 1 constructed as described above adopts a small, lightweight, and compact design that achieves non-invasiveness and high biosafety, enabling the subject to be measured and evaluated with minimal burden in a short measurement time.

[0099] Next, refer to Figure 3 The flowchart further details an example of the operation (mild cognitive impairment screening method) of the mild cognitive impairment screening device 1 constructed above.

[0100] Figure 3 This section describes one example of the processing steps performed by the mild cognitive impairment screening device 1 (steps S1 to S8 of the mild cognitive impairment screening method) (steps S2 onwards describe one example of the processing steps performed by the processor 30). As shown in the figure, in the mild cognitive impairment screening device 1 (mild cognitive impairment screening method) of this embodiment, firstly, the measurement unit 10 measures the finger tapping movements performed by the subject (measurement step S1). Specifically, the measurement unit 10 magnetically measures (detects) the finger tapping movements of only the right hand and / or only the left hand and / or simultaneous finger tapping movements of both hands and / or alternating finger tapping movements of both hands using the tapping sensor 2. During this measurement process, the processor 30 acquires the detection data (measurement data) from the tapping sensor 2 (step S2).

[0101] When the measurement unit 10 measures the subject's finger tapping motion and the processor 30 receives the measurement data, the processor 30's arithmetic circuit 33 then calculates the aforementioned feature quantity based on the received measurement data (arithmetic step S3). After the arithmetic circuit 33 calculates this feature quantity, the processor 30's determination circuit 34 then performs a determination related to the MCI based on a predetermined threshold related to the calculated feature quantity.

[0102] Specifically, in this embodiment, the determination circuit 34 generates a discriminant model using machine learning incorporating artificial intelligence (model generation step S4), and uses this generated discriminant model to make a determination related to mild cognitive impairment (determination step S5). Specifically, a discriminant model is generated by inputting a pre-prepared healthy group and an MCI group into a machine learning algorithm capable of discriminating two groups, described later. Then, a determination result is output when new subject data is input. At this time, if the F1 score, which is the evaluation index in the discriminant model (machine learning model), is not above 0.7 (if "yes" is stated in step S6), an adjustment step S7 is performed to adjust the optimal value of the hyperparameters in the machine learning so that the F1 score becomes above 0.7.

[0103] As described above, when MCI determination is performed with a high accuracy of F1 score of 0.7 or higher, the determination result is displayed on the display 37 (step S8). In addition, for example, a predetermined command is input through the operation input interface 38, and the data corresponding to the command, such as the calculation data calculated by the arithmetic circuit 33 (or the measurement data output from the measurement unit 10), is also displayed on the display 37.

[0104] Figure 5 This is a table representing examples of machine learning algorithms (machine learning models; discriminative models) for distinguishing between the healthy group and the MCI group. Here, decision trees, random forests, XGBoost, LightGBM, and support vector machines (SVM) are listed as machine learning models, but are not limited to these. Figure 5 The hyperparameters and their search ranges for each machine learning model are also shown. Within each learning model, the accuracy of distinguishing between the healthy group and the MCI group varies as the inherent hyperparameters are changed. To find the optimal values ​​of the hyperparameters, the relationship between the input feature data and the hyperparameters can also be considered when determining the machine learning model. It should be noted that, for machine learning evaluation, nested cross-validation (nCV), with a double-nested cross-validation structure, can be used to simultaneously perform hyperparameter tuning and accuracy comparison.

[0105] Regarding the 248 features mentioned above, the discrimination accuracy between the MCI group and the healthy group was calculated, and the features with the top 5% F1 scores were selected, as shown below. Figure 6 The feature number (Char.No) in the illustration corresponds to the item number of the 44 features mentioned above. As shown in the figure, among the four tasks (i.e., right-hand finger tapping only, left-hand finger tapping only, simultaneous finger tapping with both hands, and alternating finger tapping with both hands), only the feature of alternating finger tapping with both hands (related to the left hand in the figure) is selected; the other three tasks are not selected. Furthermore, regarding the features, item 29 is the time it takes to close two fingers, item 31 is the number of tremors, and items 34, 36, and 40 are features related to the rhythm of the tapping; features related to the size of the movement are not selected. In this embodiment, features with an F1 score greater than 0.5 are selected for the determination of mild cognitive impairment. These are the selected features mentioned above. It should be noted that... Figure 6 In this context, recall (fitness rate) is the proportion of cases that were predicted to be positive (patients) that were actually correctly predicted, precision (reproducibility rate) is the proportion of all positive cases that were actually predicted to be positive (also known as sensitivity), and F1 is the harmonic mean of reproduction rate and fitness rate.

[0106] Figure 7 Indicating targeting Figure 5The results of hyperparameter tuning and discrimination accuracy evaluation for the five machine learning models already shown are presented in the nCV (Multi-Version) dataset. According to the results, Random Forest has the highest discrimination accuracy (F1 score). It should be noted that, in addition to hyperparameters, F1 score, fitness rate, and reproducibility rate, other factors also play a role. Figure 7 It also shows the optimal value of the hyperparameter.

[0107] As explained above, according to the configuration of this embodiment, since the massive amount of data, which consists of a wide variety of features, is processed by machine learning including artificial intelligence, no specialized knowledge is required, and MCI screening can be easily and accurately performed in a short time. Moreover, the finger tapping measurement itself can be easily performed in a non-invasive manner using a simple and compact device, thus enabling measurement and evaluation with less burden on the subject in a shorter measurement time.

[0108] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the embodiments described above and can include various modifications. For example, the embodiments described above are embodiments that have been explained in detail for the purpose of easy understanding of the present invention, and are not limited to having all the described configurations. In addition, a part of the configuration of a certain embodiment can be replaced with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of a certain embodiment. Furthermore, regarding a part of the configuration of each embodiment, other configurations can be added, deleted, or replaced.

[0109] Furthermore, the aforementioned components, functions, processing units (processors), and processing units can be partially or entirely implemented in hardware, for example, through integrated circuit design. Alternatively, the aforementioned components and functions can be implemented in software by the processor interpreting and executing programs that implement each function. The programs, tables, files, and other information implementing each function can be stored in recording devices such as memory, hard disks, SSDs (Solid State Drives), or recording media such as IC cards, SD cards, and DVDs, or stored in devices on communication networks.

[0110] Furthermore, while control lines and information lines are considered necessary in the specifications, they may not represent all control lines and information lines on the product itself. In fact, almost all components can be considered interconnected.

[0111] Explanation of reference numerals in the attached figures

[0112] 2: Tapping sensor, 10: Measuring unit, 30: Processor, 33: Calculation circuit, 34: Judgment circuit, 40: Adjustment circuit.

Claims

1. A method for screening mild cognitive impairment, characterized in that, include: The measurement procedure involves measuring the opening and closing movements of two fingers, i.e., the tapping movements of the fingers. The calculation steps, based on the measurement data obtained through the measurement steps within a predetermined time period when only the right hand's fingers are tapping and / or only the left hand's fingers are tapping and / or both hands are tapping simultaneously and / or both hands are tapping alternately, calculate predetermined characteristic quantities, and The determination step is based on the feature quantity calculated in the calculation step to make a determination related to mild cognitive impairment; In the determination step, a discrimination model is generated by machine learning that incorporates artificial intelligence, and the generated discrimination model is used to make a determination related to mild cognitive impairment.

2. The method for screening mild cognitive impairment according to claim 1, characterized in that, The mild cognitive impairment screening method further includes an adjustment step, in which the optimal values ​​of hyperparameters in the machine learning are adjusted so that the F1 score, which serves as an evaluation metric in the machine learning model, is above 0.

7.

3. The method for screening mild cognitive impairment according to claim 1, characterized in that, The machine learning algorithm used is random forest.

4. The method for screening mild cognitive impairment according to claim 1, characterized in that, The characteristic quantities include at least one of the following: the standard deviation of the contact time between the two fingers during the tapping, the difference between the intersection of the finger acceleration during the tapping and 0, the average of the time intervals between each tapping (i.e., the tapping interval), the standard deviation of the tapping interval, and the standard deviation of the local tapping interval.

5. A computer program that processes measurement data of the opening and closing movements of two fingers, i.e., finger tapping movements, characterized in that... Have the computer perform the following steps: The calculation steps, based on the measurement data obtained through the measurement steps within a predetermined time period when only the right hand's fingers are tapping and / or only the left hand's fingers are tapping and / or both hands are tapping simultaneously and / or both hands are tapping alternately, calculate predetermined characteristic quantities, and The determination step involves making a determination related to mild cognitive impairment based on the feature quantity calculated in the calculation step. In the determination step, a discrimination model is generated by machine learning that incorporates artificial intelligence, and the generated discrimination model is used to make a determination related to mild cognitive impairment.

6. The computer program according to claim 5, characterized in that, The computer is also instructed to perform an adjustment step, in which the optimal values ​​of the hyperparameters in the machine learning are adjusted so that the F1 score, which serves as an evaluation metric in the machine learning model, is above 0.

7.

7. The computer program according to claim 5, characterized in that, The machine learning algorithm used is random forest.

8. The computer program according to claim 5, characterized in that, The characteristic quantities include at least one of the following: the standard deviation of the contact time between the two fingers during the tapping, the difference between the intersection of the finger acceleration during the tapping and 0, the average of the time intervals between each tapping (i.e., the tapping interval), the standard deviation of the tapping interval, and the standard deviation of the local tapping interval.

9. A screening device for mild cognitive impairment, characterized in that, It has a measuring unit that measures the opening and closing motion of two fingers, i.e., the tapping motion of the fingers, and a processor that processes the measurement data obtained by the measuring unit. The processor has: The arithmetic circuit, based on measurement data obtained within a predetermined time period through the measurement steps during tapping with only the fingers of the right hand and / or only the fingers of the left hand and / or simultaneous tapping with both hands and / or alternating tapping with both hands, calculates a predetermined characteristic quantity, and The determination circuit makes a determination related to mild cognitive impairment based on the feature quantity calculated in the computational step; The decision circuit generates a discrimination model through machine learning that incorporates artificial intelligence, and uses the generated discrimination model to make a determination related to mild cognitive impairment.

10. The mild cognitive impairment screening device according to claim 9, characterized in that, The mild cognitive impairment screening device also has an adjustment circuit that adjusts the optimal values ​​of hyperparameters in machine learning so that the F1 score, which serves as an evaluation metric in the machine learning model, is above 0.

7.

11. The mild cognitive impairment screening device according to claim 8, characterized in that, The machine learning algorithm used is random forest.

12. The mild cognitive impairment screening device according to claim 9, characterized in that, The characteristic quantities include at least one of the following: the standard deviation of the contact time between the two fingers during the tapping, the difference between the intersection of the finger acceleration during the tapping and 0, the average of the time intervals between each tapping (i.e., the tapping interval), the standard deviation of the tapping interval, and the standard deviation of the local tapping interval.

Images